A few months back, I presented the concept of “Digital Self’” at a digital marketing conference. Digital Self, to put it succinctly, is the digital fingerprint you create knowingly and unknowingly when you engage with all things digital (Web, apps, social, connected devices, etc.).
The fundamental point of my presentation was that we give a lot more signals in the digital realm regarding ourselves than we actually realize. I also provided examples of how some high-performing marketers are starting to understand how to best leverage these signals to provide us with engaging and personalized experiences.
One of the attendees asked, “How can a marketer get to know someone’s complete digital self for effective marketing because, by its very nature, the digital signals that we give are scattered across Web sites that we visit, apps that we use, social channels that we engage with, etc., and not all the digital signals are publicly accessible?”
In response, I elaborated on how marketers do not need to have access to someone’s entire digital footprint in cyberspace to build an actionable digital fingerprint (profile) and how some of the available technologies can help.
In fact, recently published research from University of Cambridge has demonstrated that one does not need to analyze a plethora of digital signals about someone to build an actionable profile; just studying his or her Facebook Likes will do. The researchers analyzed 58,000 U.S. Facebook volunteers who shared their Likes, demographic profiles, and psychometric testing results through an app. They were then able to create statistical models that could infer sensitive personal details based only on Facebook Likes. The model was able to predict with about 60 percent accuracy whether the parents of the Facebook user separated before the user was 21 years old. The accuracy was 95 percent when the model was asked to differentiate whether the user was Caucasian-American or African-American--all this based only on the publicly visible Facebook Likes.
To be clear, the model only shows correlation and not causality, meaning that the model cannot predict a profile based on Likes with 100 percent certainty. However, the accuracy of its inference is high enough for people to take notice. Such profiling accuracy raises, very rightly, serious privacy concerns. An obvious suggestion, as written in many articles based on this research, is to tighten privacy and security controls and to self-regulate the amount of “publicly” shared digital signals.
All that is good, but what does this research mean for marketers?
This model has been created based on information shared voluntarily with the researchers. The analysis of that voluntarily shared information has enabled them to create a system that can infer sensitive and private information about others based on only Facebook Likes. It is inevitable that such models will increase in complexity, sophistication, and accuracy of inference (and prediction) in the future. Not only that, they will most likely need less and less public information about a user to infer his or her profile with reasonable accuracy.
Marketers, this is Big Data in action. In fact this is customer/consumer-related Big Data in action. Once you capture it, analyze it and make sense of it, then you might not need large volumes of data about someone to accurately infer possible reaction to your marketing messages, thereby increasing marketing effectiveness.
In essence, when done properly, Big Data analysis can help marketers infer more with less. The academics have shown us an example of how that can be done. Marketers, are you paying attention?